summaryrefslogtreecommitdiffstats
path: root/metrics
diff options
context:
space:
mode:
authorJon Bratseth <bratseth@gmail.com>2023-08-31 13:08:18 +0200
committerGitHub <noreply@github.com>2023-08-31 13:08:18 +0200
commit47ebd33df308d82e0327fe083cd8deaed9e6fb53 (patch)
tree3c3f5174afde0dd77d020ab42780344a546f407d /metrics
parentf46d67c5976e77e270002267996a559b1cb6d2c1 (diff)
parentae674d6d002ca0f99b401e1215d45d188ba81e12 (diff)
Merge pull request #27969 from vespa-engine/bjorncs/embedder-metrics
Add generic metrics for embedders
Diffstat (limited to 'metrics')
-rw-r--r--metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java5
-rw-r--r--metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java4
2 files changed, 7 insertions, 2 deletions
diff --git a/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java b/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java
index 448233162e4..ac7ecfa124a 100644
--- a/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java
+++ b/metrics/src/main/java/ai/vespa/metrics/ContainerMetrics.java
@@ -201,9 +201,10 @@ public enum ContainerMetrics implements VespaMetrics {
SERVER_TOTAL_FAILED_RESPONSE_LATENCY("serverTotalFailedResponseLatency", Unit.MILLISECOND, "Total duration for execution of failed responses"),
SERVER_TIME_TO_FIRST_BYTE("serverTimeToFirstByte", Unit.MILLISECOND, "Time from request has been received by the server until the first byte is returned to the client"),
- SERVER_STARTED_MILLIS("serverStartedMillis", Unit.MILLISECOND, "Time since the service was started");
-
+ SERVER_STARTED_MILLIS("serverStartedMillis", Unit.MILLISECOND, "Time since the service was started"),
+ EMBEDDER_LATENCY("embedder.latency", Unit.MILLISECOND, "Time spent creating an embedding"),
+ EMBEDDER_SEQUENCE_LENGTH("embedder.sequence_length", Unit.BYTE, "Size of sequence produced by tokenizer");
private final String name;
private final Unit unit;
diff --git a/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java b/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java
index 39c1e6f519f..6c4626238eb 100644
--- a/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java
+++ b/metrics/src/main/java/ai/vespa/metrics/set/VespaMetricSet.java
@@ -116,6 +116,10 @@ public class VespaMetricSet {
// Routing layer metrics
addMetric(metrics, RoutingLayerMetrics.WORKER_CONNECTIONS.max()); // Hosted Vespa only (routing layer)
+ // Embedders
+ addMetric(metrics, ContainerMetrics.EMBEDDER_LATENCY, EnumSet.of(max, sum, count));
+ addMetric(metrics, ContainerMetrics.EMBEDDER_SEQUENCE_LENGTH, EnumSet.of(max, sum, count));
+
return metrics;
}